Hey everyone
I'm still a newbie here, improving my designs - yet the thing that gets me is that I'm still at the mercy of pre-assembled crossovers. As such, I was wondering how you guys went about learning how to do your own crossovers (Such as how what each piece in the crossover is and does, how to calculate the necessary sections properly, etc.) Any resources, like books, courses and the like would be greatly appreciated
But like I said, how did you go about learning and implementing?
Cheers
I'm still a newbie here, improving my designs - yet the thing that gets me is that I'm still at the mercy of pre-assembled crossovers. As such, I was wondering how you guys went about learning how to do your own crossovers (Such as how what each piece in the crossover is and does, how to calculate the necessary sections properly, etc.) Any resources, like books, courses and the like would be greatly appreciated
But like I said, how did you go about learning and implementing?
Cheers
How does one learn to walk? Practice works miracles.
Nowadays you can practice without spending on parts.
Simulation is very good at teaching. Measuring even
better.
Nowadays you can practice without spending on parts.
Simulation is very good at teaching. Measuring even
better.
learning the formulas for inductive and capacitive reactances is the basis for most crossover design work.
loudspeaker design books from Dickason,Weems,Davis and many others are good primers.
and i believe there's a sticky on crossover design without math.. or measurement.. or something.. round here somewhere...
loudspeaker design books from Dickason,Weems,Davis and many others are good primers.
and i believe there's a sticky on crossover design without math.. or measurement.. or something.. round here somewhere...
Hi ReDress,
In my case, I was a EE designing RF and microwave filters for a living, so was already half way there. But even then, experience and intuition only go so far 🙂
The best bet is to download an easy to play with crossover modeler (such as, oh, perhaps Xsim) and see how things interact. Start assuming speaker elements just have ideal 8 ohm resistance before you start in with real speaker impedances.
Basically, the things to understand are:
The current flowing through a driver controls its output.
Resistors in series resist current at all frequencies. In parallel (I.e., across a driver) they can allow an 'end run' around it, also reducing current in the driver.
Capacitors behave similarly, but resist low frequency current more than high frequency current.
Inductors behave similarly, too, but resist high frequency current more than low.
Capacitors and inductors can interact together to operate preferentially at mid frequencies.
Though seldom discussed, crossover networks usually provide eq as importantly as they do crossover filtering.
Driver impedance curves also interact with networks as if they were from inductance or capacitance, which is one reason (of many) why universal crossovers don't really work very well.
Play with modeling somecomponents, you'll catch on.
In my case, I was a EE designing RF and microwave filters for a living, so was already half way there. But even then, experience and intuition only go so far 🙂
The best bet is to download an easy to play with crossover modeler (such as, oh, perhaps Xsim) and see how things interact. Start assuming speaker elements just have ideal 8 ohm resistance before you start in with real speaker impedances.
Basically, the things to understand are:
The current flowing through a driver controls its output.
Resistors in series resist current at all frequencies. In parallel (I.e., across a driver) they can allow an 'end run' around it, also reducing current in the driver.
Capacitors behave similarly, but resist low frequency current more than high frequency current.
Inductors behave similarly, too, but resist high frequency current more than low.
Capacitors and inductors can interact together to operate preferentially at mid frequencies.
Though seldom discussed, crossover networks usually provide eq as importantly as they do crossover filtering.
Driver impedance curves also interact with networks as if they were from inductance or capacitance, which is one reason (of many) why universal crossovers don't really work very well.
Play with modeling somecomponents, you'll catch on.
Start here: superb write-up by Allen
http://www.diyaudio.com/forums/mult...designing-crossovers-without-measurement.html
http://www.diyaudio.com/forums/mult...designing-crossovers-without-measurement.html
Last edited:
Hi,
Slowly and painfully. Allen's write up is fine except it puts
far too much emphasis on completely unnecessary zobels.
(Zobels are best avoided if not needed, due to impedance.)
The pennies dropped looking at someone who knew what
they were doing (eventually, as they say so) and can
document most of what they are doing, and explain
why, whether you agree or disagree with the choices.
Zaphaudio.com is IMO the best site for a comprehensive
tutorial on crossovers, noting that they are all parallel,
and all quite correctly 2nd or 4th order L/R acoustic.
I don't know a good site for series crossovers. The
obvious one sticks to series slavishly for no good
reason, which gives very little insight to reality.
Series is very interesting for simple 1st order stuff,
compared to parallel with real driver impedances.
rgds, sreten.
TinaTi is a great free simulator for the basics.
Also look at the free version of Basta! and
Visatons free Boxsim, speaker simulators.
Basta! has a great free manual if your into
the technicalities of how speakers work.
Slowly and painfully. Allen's write up is fine except it puts
far too much emphasis on completely unnecessary zobels.
(Zobels are best avoided if not needed, due to impedance.)
The pennies dropped looking at someone who knew what
they were doing (eventually, as they say so) and can
document most of what they are doing, and explain
why, whether you agree or disagree with the choices.
Zaphaudio.com is IMO the best site for a comprehensive
tutorial on crossovers, noting that they are all parallel,
and all quite correctly 2nd or 4th order L/R acoustic.
I don't know a good site for series crossovers. The
obvious one sticks to series slavishly for no good
reason, which gives very little insight to reality.
Series is very interesting for simple 1st order stuff,
compared to parallel with real driver impedances.
rgds, sreten.
TinaTi is a great free simulator for the basics.
Also look at the free version of Basta! and
Visatons free Boxsim, speaker simulators.
Basta! has a great free manual if your into
the technicalities of how speakers work.
Last edited:
After reading Vance Dickason's book, I wrote a suite of crossover simulation software in the eighties to pursue avoiding Zobel correction, and to control lobing through acoustic phase. Ordinarily a Zobel network can be a practical consideration and is not necessarily a bad thing.
There are two sides to crossover development.. the electrical methods, and the acoustic goals, the latter of which I learned much from what's written about on this site.
What I haven't seen mentioned here so far is learning basic electrical/electronic theory, specifically AC circuits analysis. Any old cheap textbook of that name will do, though such textbooks tend to be dry reading.
Oh, look here:
Volume II - AC : Electronics Textbook
The first 8 chapters appears to cover what you need, especially chapter 8, as a crossover is just a collection of filters, one filter for each driver.
The thing is, one doesn't stop learning, so might as well start somewhere... anywhere... everywhere.
Always keep in mind that the crossover design for any particular set of drivers not only involves coils, capacitors and resistors but also baffle shape, its size and driver placement on it too.
A successful loudspeaker design is a refinement of all these things even down to the timber and any exotic materials used in its manufacture. they all contribute in some way to the art of making beautiful loudspeakers. It's an art form in itself.
C.M
A successful loudspeaker design is a refinement of all these things even down to the timber and any exotic materials used in its manufacture. they all contribute in some way to the art of making beautiful loudspeakers. It's an art form in itself.
C.M
LR2 and LR4 are only 2 types of designing mantra, and lots more are worthwhile. It really depends on the drivers used. LR4 is basically the 'easiest' way of doing it in most cases.
Andy G used to have a website devoted to 1st order SXO's. I have a primer here that basically compiled the thoughts of Andy G, Clayton Oxendine, and some of my own thrown in. I feel they are just another way of doing it, and aren't necessarily 'better'; just different. You can use all sorts of notches and tricks, and even use 'difficult' drivers just as you can in parallel topologies, but usage is different.
In terms of where I started for the OP....
Back in 1999 I found a copy of Ray Alden's 'Building Speaker Systems' from Radio Shack- and it just snowballed. It helped me with the basic equations and getting used to about what values would theoretically yield what results in lots of cases. Then- real world drivers and applications came into play that changed a lot of those thought processes. Acoustics and how sound ripples in a room helped with baffle-step and diffraction, as well as room interaction. The original theory in math was still sound to getting close to where I wanted to be approximately, and I still use that by educated guesses and practice rather than doing them math every time. You get accustomed to approximately what will work. About 2006 gave way to my WT2 and Jeff B's PCD/RM suite of excel tools, and my monster was unleashed. I had the circuit knowledge, acoustic applications, etc, and this gave the path to using it all with much less trial and error with very accurate results if you were as accurate as possible in your setting up your files.
There is still an 'art' to designing a xover. Everybody actually does it differently, and seldom will one approach be the only method that works well. Only a few people I've personally corresponded with seem to have the same methods as I use, and designs that are voiced similarly as well. BUT- they really studied my methods to get their methods of practice down to where it works for them.
It's a rewarding hobby, and I enjoy it a lot!
Happy building,
Wolf
Get a copy of Vance Dickasons "Loudspeaker Recipes" and start reading. Even if the text may be hard to understand in the beginning, the various graphs in the book will show you you what in-box-raw-drivers measurements look like and, more important, how filters must interact (or complement) with the drivers in order to obtain the desired acoustic output a.k.a the acoustic target.
Have fun,
Eelco
Have fun,
Eelco
As mentioned, acoustics do come into play with lsp-filters. Not only things like baffle step but also directivity, 'loading' a conventional driver (as in the spring and damping mechanism of the enclosure) and the implications on electrical filtering to name a few aspects.
Knowledge of room interaction, first bounces and understanding of sound fields in general will be helpful. But just start, get on and be willing to learn. Any understanding of how our ears work and what we can and cannot perceive is useful too.
Don't get too confused on topics like component quality, a lot of audiophiles are and will always be producing distracting noise on that. Focus on the electrical-acoustical interaction of elements in the sound path to your ears.
Knowledge of room interaction, first bounces and understanding of sound fields in general will be helpful. But just start, get on and be willing to learn. Any understanding of how our ears work and what we can and cannot perceive is useful too.
Don't get too confused on topics like component quality, a lot of audiophiles are and will always be producing distracting noise on that. Focus on the electrical-acoustical interaction of elements in the sound path to your ears.
I started by jumping in the deep end! I measured the drivers acoustically and their impedance and started to simulate in speaker workshop.
Actually the very first thing I did was just put a cap on the tweeter and let the mids run full range. This sounded surprisingly good for some music and awful with other music.
I simulated a lot then built what I simulated. I listened to that for about a year, then I read some more here on diyAudio and decided to start simulating again. This resulted in a much better simulation, which I then implemented and have been listening to ever since.
I recently learnt a few more tricks, and may well do another revision of the crossover.
One very important thing for me was making sure that the realized results matched the simulations. My last revision of the crossover was built as simulated and has not been changed.
I chose drivers that I thought were going to be relatively easy to work with. Unfortunately they had some response anomalies that when combined with the baffle step were quite a challenge do deal with, however in the end this taught me more. I would however recommend starting out with drivers with smooth roll offs and no nasty response anomalies.
With respect to Sreten's comment on zobels, I designed without them too, (I did try one but did not like the result). BUT in the specific context of Allen's sticky thread I would say that they are almost certainly essential, as without measurements and simulation I think that designing without them is likely to be very hit and miss.
Tony.
Actually the very first thing I did was just put a cap on the tweeter and let the mids run full range. This sounded surprisingly good for some music and awful with other music.
I simulated a lot then built what I simulated. I listened to that for about a year, then I read some more here on diyAudio and decided to start simulating again. This resulted in a much better simulation, which I then implemented and have been listening to ever since.
I recently learnt a few more tricks, and may well do another revision of the crossover.
One very important thing for me was making sure that the realized results matched the simulations. My last revision of the crossover was built as simulated and has not been changed.
I chose drivers that I thought were going to be relatively easy to work with. Unfortunately they had some response anomalies that when combined with the baffle step were quite a challenge do deal with, however in the end this taught me more. I would however recommend starting out with drivers with smooth roll offs and no nasty response anomalies.
With respect to Sreten's comment on zobels, I designed without them too, (I did try one but did not like the result). BUT in the specific context of Allen's sticky thread I would say that they are almost certainly essential, as without measurements and simulation I think that designing without them is likely to be very hit and miss.
Tony.
I'd challenge others to reverse engineer a known good filter to incorporate a textbook impedance correction so that the response is otherwise identical.. then to show a significant difference.
I started by doing it ears-only. Then with additional high-school Physics formula. Then with additional simulation. Then with additional measurement. Then gradually go back to ears-only 😀
Crossovers are simple circuits, with lots of experience you will know what happen when you make changes here and there, and ears (when reliable) are very good measurement tool.
@ Jay: Do you really think the designers at Genelec or JBL. KEF and others think that ears are a very good measurement tool? No they do not.
Wintermute is fully right, You measure first, you simulate, you build, you measure again, you listen and and then finally the ears are used for a bit of tweaking, mostly in the form of lowering or raising tweeter level a decibel or so. And if you do not like the final result, you start all ver again but then with a different filter topology
BTW without a proper simulation tool the whole exercise is pointless
Good luck ,
Eelco
Wintermute is fully right, You measure first, you simulate, you build, you measure again, you listen and and then finally the ears are used for a bit of tweaking, mostly in the form of lowering or raising tweeter level a decibel or so. And if you do not like the final result, you start all ver again but then with a different filter topology
BTW without a proper simulation tool the whole exercise is pointless
Good luck ,
Eelco
Make that proper measurement equipment (and knowing how to use it). Back in the eighties I had no access to sim software. But yet with a Neutrik set and a closet full of filter components I mostly succeeded pretty well.
- Status
- Not open for further replies.